Q9550 system Power Consumption Test

[Cyber-Mav]

though i would run a few tests and share the results with everyone on here. i purchased a power meter: http://www.maplin.co.uk/Module.aspx?ModuleNo=223573

and gave it a whirl on my system:

Media PC Powermeter tests:

Power Meter Used was from Maplin's, a Nikkai Power code N67FU


system:

Q9550@3.4ghz (stock vid 1.2500v used) (speed step enabled) (idle vid = 1.1000v)
4x1gb corsair xms3 dhx (1600C7) @ 1280mhz 6-6-6-18 1.80v
Gigabyte P35T-DQ6 rev1.0
cpu settings 400 x 8.5 - speedsteps to 400 x 6.0 Load Line Calibration disabled
core contact freezer + scythe sflex 1600 (full speed)
2 x scythe sflex 1200 (full speed)
gainward gtx260 216 GS 55nm (clocked to 700core, 1200ram, 1440shader)
wd black 1tb
pioneer dvr216D
corsair HX520w psu

(all are peak readings obtained)

System Powered Off = 2.4w


Idle in windows doing nothing, no HDD activity, speedstep clocked to 2.4ghz = 108.2w


LINX (linpak) stress test speedstep gone to 3.4ghz = 222.7w


GPU gone to 3d clocks when occt loaded up but no stress testing = 128.2w


OCCT GPU test started full screen (1360x768 rez 52fps average) = 271.8w


OCCT GPU + LINX run at the same time = 383.7w



as the results show, going from idle to max gpu load is around 170w difference, if i was running sli with 2 cards add another 170w and that makes it nearly 554w + a bit more since gpu probably uses more than 170w since when idle its probably sucking up a few more watts. take into account 80% psu efficiency and iv got around 445w actual pull from the psu, pull from the mains is always higher due to efficiency loss. so its possible a hx520w psu can power sli rigs.

idle consumption i thought was impressive general web browsing etc didnt even get to 110w.

by the way people this is done on uk 240v mains.

i used linx (linpak) and occt gpu tests since they are the most extreme tests out there for gpu and cpu.

peak power pulled from the wall socket was as shown above 383.7w the pc itself is pulling from the powersupply taking into account 80% efficiency
= 383.7 x 0.80 which = 307w pc is pulling from the power supply itself.


hope this info helps out others looking for power supplies.

[Conquistador SW]

Excellent post! We should make a list ore something with common configurations so people can see what kind of power supply they need.

I have a similar setup and see also see a 400W max load when running prime 95 and 3Dmark together.
(q9450 3.6Ghz 1.28v / GTX 260 192 640/2300 / for rest see sig)

The only thing you got wrong is how the 80% efficiency works. If you see a 400W pull from the socket (What your wattage meter measures) that resembles the 100% pull. Your PSU will only deliver 80% of that to the system, so the system itself will only pull about 320W from the PSU. From this you can see that you actually don’t need a high wattage PSU to power a somewhat “high end” system.

[Cyber-Mav]

yes, i mentioned the 445w pull from the psu IF i had an sli configuration. the wall socket would have seen more.

also prime is a lot lower on cpu power pulling compared to linx, 27w lower prime was compared to linpak.

real test should have been to run hdtune on the hard drive and have me pressing the eject button loads of times on the dvd drive to get a bit more wattage up. lol

[Hidetaka]

Could you check the power draw while playing some demanding games? Most people don't run OCCT and Linx daily while results from running Crysis/FC2 or some other new game could show actual cost of running a high end gaming rig.

[Cyber-Mav]

yep i will run some gaming power tests at somepoint. the system configuration has changed since i now have 8gb ram in it and use a 620w corsair psu so i will most likley put a new thread up with all the info in it.

[Jamesrt2004]

theres someone on here with i7 @ 4.2+HT and dual 4870's etc... on his rig and he uses a 520hx

so good results =)

(also remeber effieciency will be based only at "normal" load, so idle/watchign a film or something) so under load its probably a tad worse (although I could not say what I could only hazard a guess which isn't very informative)